Automatically generating personalized user interfaces with Supple
Download 5.78 Mb. View original pdf
|
| Fig. 35. Participant completion times. Both motor-impaired and able-bodied participants were fastest with the ability-based interfaces. The baseline interfaces were slowest to use. Error bars show standard error. using regression with an exponential distribution [84]. Subjective Likert scale responses were analyzed with ordinal logistic regression [90], and subjective ranking data with the Friedman non-parametric test. For all measures, additional pairwise comparisons between interface variants were done using a Wilcoxon Signed Rank test with Holm’s sequential Bonferroni procedure [37]. 8.6. Results 8.6.1. Adjustment of data We excluded 2 / 765 trial sets for two different motor-impaired participants, one due to an error in logging, and one because the participant got distracted for an extended period of time by an unrelated event. 8.6.2. Completion times Both Impairment (F 1 , 15 = 28 . 14, p < . 0001) and Interface variant (F 2 , 674 = 228 . 30, p < . 0001) had a significant effect on the total task completion time. Motor-impaired users needed on averages to complete atrial set while able- bodied participants needed only 18.2 s. The ability-based interfaces were fastest to uses, followed by preference- based (26.0 sand baselines (28.2 s. A significant interaction between Impairment and Interface variant (F 2 , 674 = 6 . 44, p < . 01) indicates that the two groups saw different gains over the baselines from the two personalized interface variants. 944 K.Z. Gajos et al. / Artificial Intelligence 174 (2010) 910–950 Download 5.78 Mb. Share with your friends:
|